feat(python/sedonadb): add DataFrame.group_by + GroupedDataFrame.agg

[jiayuasu]

Grouped aggregation on top of the registry-driven function dispatch (#885) and the global-aggregation binding (#887). Completes the aggregation track of Phase P2 (#791).

API

df.group_by("k").agg(total=sd.funcs.sum(sd.col("v")))

df.group_by("k1", "k2").agg(
    sd.funcs.sum(col("x")).alias("sum_x"),
    n=sd.funcs.count(col("y")),
)

df.group_by(col("x") + col("y")).agg(...)         # computed-Expr group key
df.group_by(col("k"), "other_key").agg(...)        # mixed str / Expr

• df.group_by(*keys) — varargs of str | Expr. Strings auto-promote to col(name) (same pattern as sort). Empty keys → ValueError; non-str/non-Expr → TypeError.
• Returns a new GroupedDataFrame — thin holder for the parent df + resolved group exprs.
• GroupedDataFrame.agg(*exprs, **named_exprs) — same signature as DataFrame.agg, including the kwargs-as-alias shorthand.

Implementation

Pure Python — no Rust changes. The InternalDataFrame::aggregate(group_exprs, agg_exprs) binding from #887 already handles the grouped case; this PR just populates group_exprs when constructing the aggregation. The GroupedDataFrame intermediate is kept minimal (one method beyond __init__) so it stays a clean place to add convenience aggregates (count, size, etc.) later without polluting DataFrame.

Test plan

12 tests in tests/expr/test_dataframe_group_by.py:

• Positive: single-key string; multi-key strings; col(k) Expr key; computed-Expr key (col("x") + col("y")); mixed str / Expr; positional + kwarg agg in one call.
• Lazy return: df.group_by("k") → GroupedDataFrame; .agg(...) → DataFrame.
• Errors: empty group_by() → ValueError; bad-type key → TypeError; empty .agg() → ValueError; non-Expr agg → TypeError.

Output assertions use pd.testing.assert_frame_equal after sorting (group-by output ordering isn't guaranteed).

Local: 12 unit + 23 doctests + ruff format + ruff check all clean.

[paleolimbot]

Thank you!

[paleolimbot]

Can you inline the sort into the tests? .group_by().agg().sort() would be about as compact and make it easier to rearrange the tests later.

[jiayuasu]

Inlined in 6a20692 — _sorted helper is gone; each test that needed deterministic group ordering now chains .sort(...).to_pandas().reset_index(drop=True) directly.

[paleolimbot]

Suggested change

	Produced by `DataFrame.group_by(...)`. The only public method is
	`agg(...)`, which runs the aggregation and returns a new
	`DataFrame` with one row per group. The class exists as a step in
	the chain so that future convenience aggregates (e.g. `count()`,
	`size()`) can land here without polluting `DataFrame`.
	Produced by `DataFrame.group_by(...)`. The class exists as a step in
	the chain to simplify aggregation expressions.

[jiayuasu]

Applied verbatim in 6a20692.

[paleolimbot]

Suggested change

	# Tests for DataFrame.group_by(*keys).agg(*exprs, **named_exprs).
	# Aggregate exprs come from `con.funcs.<name>(args)` via the function
	# registry (#885); the Rust binding is shared with `df.agg`.

[jiayuasu]

Dropped in 39bdba4 — module-level meta comment removed (deleting just the trailing blank would have left it adjacent to imports); five .reset_index(drop=True) calls also gone since to_pandas() of a fresh sedona DataFrame already returns a default RangeIndex.

[paleolimbot]

I don't think these do anything since the fresh pandas df is not grouped? (With apologies if they're needed)

Suggested change

.reset_index(drop=True)

[jiayuasu]

Dropped in 39bdba4 — module-level meta comment removed (deleting just the trailing blank would have left it adjacent to imports); five .reset_index(drop=True) calls also gone since to_pandas() of a fresh sedona DataFrame already returns a default RangeIndex.

[paleolimbot]

Suggested change

.reset_index(drop=True)

[jiayuasu]

Dropped in 39bdba4 — module-level meta comment removed (deleting just the trailing blank would have left it adjacent to imports); five .reset_index(drop=True) calls also gone since to_pandas() of a fresh sedona DataFrame already returns a default RangeIndex.

[paleolimbot]

Suggested change

.reset_index(drop=True)

[jiayuasu]

Dropped in 39bdba4 — module-level meta comment removed (deleting just the trailing blank would have left it adjacent to imports); five .reset_index(drop=True) calls also gone since to_pandas() of a fresh sedona DataFrame already returns a default RangeIndex.

[paleolimbot]

Suggested change

.reset_index(drop=True)

[jiayuasu]

Dropped in 39bdba4 — module-level meta comment removed (deleting just the trailing blank would have left it adjacent to imports); five .reset_index(drop=True) calls also gone since to_pandas() of a fresh sedona DataFrame already returns a default RangeIndex.

[paleolimbot]

Suggested change

.reset_index(drop=True)

[jiayuasu]

Dropped in 39bdba4 — module-level meta comment removed (deleting just the trailing blank would have left it adjacent to imports); five .reset_index(drop=True) calls also gone since to_pandas() of a fresh sedona DataFrame already returns a default RangeIndex.